I'll preface all this by saying that I could be an idiot and I'm probably completely wrong. I only took my physics education to A level and my awarenss on this topic is introductory at best. I just figured I should bring it up and someone educated can correct me or show me whats up -- but I wanted to bring up this topic. This really only comes out of my experience in my hobby, blacksmithing. I think a user like Goosah as a mechanical engineer or someone who knows their physics beyond A level could offer something far more educated So a lot of people criticise BeamNG of being all “jelly” and yes, I hate to admit it but there could be something to it. So the vehicles are the right sort of strength, shown time and time again that in BeamNG you get a nice approximation of a crash that’s fairly accurate. But then I see things like this Well, okay that might give the jelly theorists some validation. We know from crash test footage that the covet is the right strength in an accident yet in real life my little Suzuki’s frame doesn’t rattle this badly when I bump the curb a little hard trying to park. The Covet’s does and it’s likely nothing that Gabester can help. Car bodies aren’t held together by solid steel bars because we want cars to be light to get good fuel economy and save on material costs. In unibody construction we find that cars are thin metal sheets stamped into shapes and welded together into a very rigid structure, which is why I can show you a picture of a cross section taken out of a car’s roof and you can see daylight through the actual structural parts of it Now maybe you're familliar with corrugated steel, it's that wavy sheet metal you build sheds out of. Why is it corrugated? it makes the steel more rigid without it being any thicker. What you generally find, and anyone with experience using this stuff can back me up, is that corrugated steel is very flimsy in one direction and quite rigid in the direction perpendicular to the corrugations in the steel. This means to me that corrugations make the sheet steel stronger, so is the solution simply to make the beams in each car stronger? No, it isn't because if we make them stronger then yes the car is more rigid but it won't be accurate in a crash anymore. now remember, I said hooke's law is to blame. How are beams simulated in BeamNG? From what i can gather, the beams are basically springs that increase and decrease in length according to hooke's law. They have an elastic limit, and then they permanently deform after that limit and break when they deform too much. so I have a feeling that if I plot the stress/strain graph for a beam in BeamNG I might find that it will very closely follow this curve Now I know from my hobby blacksmithing and spending countless years shaping and bending iron that In terms of physical deformation the failure of a solid metal bar is quite steady, which means when shaping metal I don’t keep breaking my wrists. But how do corrugated structures fail? Well I made an unscientific little video what I did in this experiement was take a piece of tooling foil and bend it into a little beam, you can see it in it's broken state under the video there. This is a beam with a consistent cross section. As I steadily lowered that bit of steel onto the foil beam I made it initially took the weight quite well and flexed under the load as you saw. But as I continued to rest more and more weight as steadily as I could on that beam there’s a point where it suddenly and catastrophically just “snapped”. Car bodies in real life do this too, the passenger cell of a vehicle is VERY rigid until you overload it and then it’ll just “snap” like that beam. Typically for a car once you compromise the structure then it’ll never be as strong again – half the reason a cut-n-weld is a very dangerous thing on the road. Again, I blacksmith. If a bar of solid steel behaved like this when bending it I would have broken wrists because folded metal structures deform suddenly past their elastic limit, not steadily. This is why in order to bend a copper tube without kinking it I need to fill it with sand. This means to me that the two things are distinct from one another and when you want to simulate a tube in BeamNG, it should behave differently to an iron bar. If you want to simulate a car body more accurately in BeamNG then it might be something to ponder about. So let me take the example of a car’s A pillar to think about. The red dots are obviously nodes, and in blue are beams that represent the car’s shape. I’m only considering the A pillar, and the green beam is the one that will simulate the A pillar’s structural rigidity. It’s a rigidifying beam. Currently in BeamNG this green beam deforms steadily according to hooke’s law but the A pillar of a car is essentially a bent metal tube. when it fails it should give suddenly once the load becomes too great, yet as things are now the structure will deform steadily under load. I want it to be rigid and then “snap” like in that video. So I could set it to break, but that’s no good because if I did that then the car would become soup. So what needs to happen is when this beam breaks it simply becomes far less rigid than before it broke rather than stopping existing. I’m not sure there exists a beam type that behaves this way, but if not it could be an argument for a new beam type. If there does exist a beam that behaves like this, take it as a little suggestion for content creators to be aware of this sort of thing. I dunno. Lemme know what you think
You are actually well on track Structures that buckle have a distinctly non linear deformation, which we are approximating with generally linear beams, however, many of our nb structures are actually buckling themselves (roofs, pillars, etc). We could further replicate this with extra beam properties, with something similar to how bounded beams have different spring and damping beyond specified bounds. Alternatively, as you said, we could add more brittle beams that could break and leave a weakened structure. Another way would be to increase the nb density to allow more detailed buckling behavior of the nb structure itself (probably the most elegant solution). The problem with these ideas is that I don't see a way to accomplish it without increasing the nb "budget" of the vehicles quite a bit. Not to mention the extra overhead in jbeam tuning. Personally I think if we are to increase the nb budget it should be devoted to adding more functionality, and to make the environment more interactive. The video of the covet bumping the barrier is interesting. The beams in this case are not deforming at all. It may be we need to add some structure with low deform values to help with rigidity of the cowl/a pillar. - - - Updated - - - Oh one more thing to add. There are cases where buckling deformation is not so sudden. Crumple zones are designed to be very linear. Crumple zones are made of panels or hollow structures that do not buckle out of the way but folds on themselves in a series of small buckles, like an accordion. Crushing a rocker panel or buckling a pillar is much as you describe, but certain collisions and certain areas of the car have a more linear deformation
Isn't the Beamspring value the one most responsible for the "jelly" effect? I've messed around with that in slo mo and strengthening those values appears to lessen the jelly effect there by affecting how and when the area starts it's collapse.
Its partially beamSpring. In most cases beamSpring is already on the edge of stability. Further improvements have to come from more triangulation or rigidifier beams.
Thankyou for the rounded response, glad to see I'm on track and it's good to hear some dev positions on this. You wouldn't have some pointers on where I could read more on this topic? it would be nice if I knew what I was talking about some more! Maybe advanced beam properties would be a good way to handle it that wouldn't rack up the node/beam complexity. But something that might simply take advantage of existing code could do it. There exists a "broken" state for beams in this engine. So already in BeamNG you have a beam is defined like this {"breakGroup":"frame"}, {"beamSpring":1251000,"beamDamp":250}, {"beamDeform":16000,"beamStrength":24000}, ["f1rr","f1r"], Taken from the wiki. So what I have here is beam f1rr,f1r with a finite strength value. So if it's deformed beyond it's strength threshold the beam is "broken" or deleted. What if I added a set of values with Alt in the name so that BeamAltstrength is something that says "If deformed above the threshold defined by BeamAltStrength, change the beam's physical properties to Altspring and AltDamp etc" {"breakGroup":"frame"}, {"beamSpring":1251000,"beamDamp":250}, {"beamDeform":16000,"beamAltStrength":24000}, {"BeamAltSpring:90000,"BeamAltDamp":200}, {"BeamAltDeform":10000, "BeamStrength":1200"} ["f1rr","f1r"], My new beam type would look like that, but I don't see it using many more resources than a normal beam since it should work the same way. It goes through whatever check is needed to break the beam, but instead of breaking it is switched over to being simulated with the Alt values, and if further deformed then the beam is then broken with the regular "beamStrength" threshold. I hope I'm clear. But wouldn't something like that work?
Essentially you're creating a form of bounded beam, and those have a higher calculation cost than a regular beam. We only use them for specific cases like limiters and suspension springs.
